PRISM/INVESTOR BRIEFING
01 / DIRECT INDEXING · U.S. TAXABLE ACCOUNTS

Personalized portfolios. Tax-aware by design.

Direct indexing: own the stocks, manage individual holdings, and seek index-like exposure. An ETF opens into individually owned stock bubbles. Holdings move and resize continuously. Multiple eligible loss-bearing lots are sold and different securities replace their exposure after review. Steel index and teal portfolio lines fluctuate together, with the latest segment magnified. All data is illustrative, not a performance forecast. Playback controls are below the image. FUND SHARES Index ETF / mutual fund DIRECT INDEXING Your stocks. Your rules. Loss TAX-LOSS HARVESTING Stock-level choices Exclude Adjust weights ILLUSTRATIVE · $1B PORTFOLIO $10M Eligible losses × 20% $2M Potential tax offset* *Assumed rate · sufficient gains · not investment profit Seek index-like exposure INDEX EXPOSURE Follow the index LATEST WINDOW Index Portfolio INDEX & PORTFOLIO RETURNS Illustrative paths · before tax & fees Monthly return differences · their variability is tracking error COLOR = SECTOR · SIZE ≈ WEIGHT FACTOR ALIGNMENT Index target WASH-SALE CHECK Same / substantially identical −30 daysSale+30 days Screen linked accounts
Illustrative · before taxes & fees · *replacement eligibility assumed after review
SOFTWARE-REVENUE SCENARIO

Contracts, not AUM.

Enter assumptions. This is a transparent model, not a researched TAM/SAM/SOM or a forecast.

SCENARIO REVENUE TAMNot measured
SCENARIO REVENUE SAMNot measured
SCENARIO ANNUAL SOMNot measured

Illustrative contract arithmetic only. The same portfolio may use multiple workflows; do not count its AUM repeatedly. No tax-saving or investment-value estimator is included.
RESEARCH & ASSUMPTIONS · 16 SEPTEMBER 2026

The commercial model.

USD throughout. Proposed pricing and financing; public evidence is identified separately.

Strategy: own the decision layer

PRISM’s own CPU/GPU engines remain the measured foundation. Proposed adapters would let qualified Gurobi, NVIDIA cuOpt and future QPUs serve selected workloads behind one client-facing decision service. Integrations, redistribution rights and per-model qualification must be established; the deck does not claim existing integrations or a cuOpt benchmark. The intended moat is permissioned workflow integration, client-rule history, auditable validation records, measured service reliability and reusable onboarding. These require adoption to become defensible assets.

Expansion gates: first win direct indexing and rebalancing at large US platforms; expand into wealth and asset management after repeatable deployments and renewals; enter energy or other deadline-driven domains after separate model and buyer validation; explore error-corrected quantum finance around 2030+ only if hardware availability and end-to-end economic benefit justify it. This fourth stage is not dependent on launching energy first. Avoid a fixed market-share or trillion-dollar valuation promise. At an illustrative 20× annual-revenue multiple, $1T equity value requires $50B annual revenue; even the $147T global-AUM ceiling at 1 bp yields only $14.7B hypothetical annual software revenue before market share and costs. A trillion-dollar outcome is therefore not justified by this financial-market model.

Quantum: explicit, conditional economics

Base quantum revenue and valuation premium: $0. Proposed optional software module: $25K per platform annually, excluding QPU usage and integration. Same 20 candidates × $25K = $500K maximum cohort option if all qualify, not an additional asset market or base SAM. One module at month 36 adds $25K to $750K base ARR = $775K; this is conditional on useful quantum hardware, measured economic advantage and a paying customer. No separately sized quantum TAM is claimed.

The $2M seed, $8M pre-money, 20% dilution and 24-month runway remain proposed terms. Faster benchmark results do not mechanically increase valuation. Within the existing $1M engineering allocation, earmark $100K for quantum research; within the existing $160K compute/data allocation, earmark $20K for quantum experiments. Thus quantum is $120K, or 6% of the raise, and the remaining $1.88M covers all other items including the $200K reserve. The total operating budget remains $1.8M, or $75K per month. No extra headcount or QPU-hardware purchase is assumed.

Refreshed public research · 17 September 2026

Gurobi: commercial quote-based licensing. NVIDIA cuOpt: open-source Apache 2.0 engine; infrastructure and enterprise support are separate. A GPU solver alone cannot justify premium platform pricing.

MOSEK: PTS-NODE $9,000 perpetual, $2,250 annual maintenance. IBM CPLEX: starts at $320 per authorized user per month; not a production platform quote.

Cerulli: $864.3B DI assets at year-end 2024, retained as historical context; the current model instead uses MMI–Cerulli Q1 2026 $1.2T. BCG: $147T global AUM at year-end 2025, an ecosystem ceiling. IBM roadmap: targets fault-tolerant hardware in 2029; it does not establish PRISM advantage, commercial availability on required terms, or a finance TAM.

What the pricing research supports

Category / providerObserved basisMeaning for PRISM
MOSEK · PTS-NODE$9,000 perpetual + optional $2,250 annual maintenanceBase solver on one machine. A $150K platform licence must earn its premium through integration and operational value.
IBM CPLEXFrom $320 / user / month, development only; production quoted$3,840 annualised developer price is not an enterprise deployment price.
Gurobi / FICO XpressGurobi requests a quote. No standard Xpress production price disclosed on the reviewed page.No invented list prices or asserted licence savings.
Vestmark / Smartleaf / MSCI BarraVestmark: account / asset-linked pricing in its commissioned study. Smartleaf: price undisclosed in reviewed material. No verified Barra rate card obtained.Portfolio platforms are closer in workflow, broader in scope. Their prices do not establish PRISM willingness to pay.
Wealthfront S&P 500 Direct0.09% / year, or 9 bps, managed serviceContext only. $1.5B at 9 bps = $1.35M management revenue; a $150K PRISM fee would consume 11.1% before other costs. This is not a partnership or a like-for-like price comparison.

Proposed offer and economic test

$25K one-off, eight-week paid pilot. Annual licence = $100K platform + $5 × contracted active-account capacity. Reference: 10K accounts = $150K; 50K = $350K; 100K = $600K. One bounded equity workflow and an agreed solve cadence. Customer VPC compute, market data, third-party licences and implementation are separate. Support, account definition, burst volume and service levels must be scoped after the pilot; the benchmark is not a fleet-capacity guarantee.

At 10K accounts of an assumed $150K each, the eligible book is $1.5B. The proposed $150K licence is equivalent to 1 bp of that book. At a $1B book it is 1.5 bps. This is a normalisation, not AUM billing. Target a customer-validated annual operating benefit of at least $450K (3× the $150K fee), after incremental compute / integration costs; no saving or tax-alpha claim is established. If the customer only needs a faster generic solver, this premium may not be justified.

TAM / SAM / SOM: explicit scenarios

Direct-indexing TAM proxy: MMI–Cerulli’s $1.2T Q1 2026 manager-traded DI assets × assumed 1 bp = $120M annual software potential. At 0.5–2 bps: $60–240M. This is not measured software spending, measured software spending, or a company-wide global TAM. Pricing is account-based; the yield is a separate top-down cross-check. Its precision comes from arithmetic, not market certainty.

Initial candidate SAM: 23 Tier 1 spreadsheet records − 1 Envestnet/Tamarac duplicate − 2 input-provider records (MSCI/Axioma, FactSet) = 20 candidate organisations. 20 × assumed $150K average annual contract = $3M. This is a named target-cohort scenario, not an exhaustive or validated SAM; each organisation still needs technical, budget and procurement qualification. It is not a measured subset of Cerulli's asset pool.

36-month SOM: 5 contracts × $150K = $750K ARR, requiring 25% of the initial cohort to convert. Downside 3 = $450K; upside 8 = $1.2M. No pilot fees, implementation fees, asset pools or modules are added again. These are goals, not probability-weighted forecasts or pipeline commitments.

Seed: $2M for a 24-month operating plan

Proposed priced round: $8M pre-money + $2M new cash = $10M post-money; 20% new-money ownership before an option-pool top-up or convertibles. This is a negotiating scenario, not an appraisal. Carta reports roughly 19–20% early-stage dilution and a $24M Q4-2025 US seed median post-money; those samples are context, not a French / European comp or evidence that PRISM is worth a particular amount.

Engineering $1M: founder allowance $200K over 24 months, two engineers at $160K fully loaded / year each ($640K), $160K specialist / QA allowance. Deployment $400K: forward-deployment role $250K over 24 months, integrations $100K, travel / selling $50K. Compute & data $160K. Security, legal & operations $240K (security $100K, legal $60K, operating / accounting / insurance $80K). Reserve $200K. All are proposed USD costs, not actual compensation or vendor quotes.

Operating spend $1.8M / 24 = $75K average / month; $200K reserve is additional. No revenue, tax credits, current cash or cloud grants subsidise runway. With the reserve preserved, a 25% cost overrun reduces runway to 19.2 months. FX, payroll location and hiring dates require budget validation. Existing bootstrap documents excluded salaries and do not establish this funded-team runway.

Targets from funding close: month 6 — reproducible customer-workload validation and first paid pilots; month 12 — eight cumulative paid pilots and first production contract; month 24 — three annual production contracts ($450K ARR) plus referenceable operating evidence. Pilot receipts are nonrecurring and excluded from ARR. Month-36 SOM is a later goal and may require further financing.

Sources

  1. P1 · MOSEK commercial pricing ↗

    Published USD prices, effective September 1, 2025. PTS-NODE base: $9,000 perpetual plus optional $2,250 annual maintenance. A solver licence, not a managed portfolio platform.

  2. P2 · IBM CPLEX pricing ↗

    Developer subscription starts at $320 per authorised user per month; development only. Production deployment is quoted.

  3. P3 · Gurobi licensing ↗

    Commercial pricing is requested by quote; workstation, server, cloud and container licensing. No public standard production price used.

  4. P4 · VestmarkONE economic-impact study ↗

    Vendor-commissioned Forrester study describes account / asset-based pricing. Composite solution costs include implementation and change management; not treated as a software rate card.

  5. P5 · Smartleaf portfolio management ↗

    Portfolio management and tax-customisation scope. No standard software price disclosed in the reviewed official materials.

  6. P6 · Wealthfront S&P 500 Direct ↗

    Published 0.09% annual management fee (9 basis points). Managed investment service; contextual economics, not a like-for-like software competitor price.

  7. P7 · Cerulli direct-indexing assets ↗

    April 10, 2025 release: $864.3B direct-indexing assets at year-end 2024. Historical asset base, not a current revenue forecast. The software yield is our assumption.

  8. P8 · Carta early-stage financing context ↗

    March 5, 2026: Q4 2025 primary seed median post-money valuation $24M; seed / Series A dilution approximately 19–20%. US Carta sample, not a PRISM valuation or a local market comp.

  9. P9 · FICO Xpress licensing ↗

    Commercial licensing available; no standard production dollar price disclosed on the reviewed official page.

  10. P10 · Internal target cohort and proposed model

    PRISM_Target_Company_Pipeline_Tier_1_3.xlsx: 23 Tier 1 records. Combine Envestnet / Tamarac, exclude MSCI / Axioma and FactSet input-provider records: 20 candidate organisations. Fit, budget and procurement remain assumptions, not customer commitments.
DIRECT INDEXING / SOURCES & MODEL

Behind the animation.

Ownership

Index investing describes a strategy. An index ETF or index mutual fund pools investors’ money; the investor owns fund shares. Direct indexing holds individual securities in a separate portfolio, allowing stock-level choices. Funds can also be tax efficient, and their shares can be harvested when those shares have a loss. ETFs usually trade intraday; mutual funds generally transact at end-of-day NAV.

SEC Investor.gov: fund structures ↗Fidelity: direct ownership and tradeoffs ↗

Wash-sale screen

U.S. wash-sale rules can disallow a loss when substantially identical securities are acquired during the 30 days before or after the sale, including the sale day. Screen linked taxable accounts, spouse activity, automatic reinvestment, relevant options and IRA/Roth IRA acquisitions. Broker reporting may not capture every cross-account event. Taxable-account wash losses generally adjust replacement basis; IRA replacement purchases have a different, potentially permanent result. A different ticker or similar sector alone does not establish eligibility. Each replacement in the scene is assumed eligible after review; the animation cannot establish real-world eligibility.

IRS Publication 550: wash sales and tax lots ↗

Losses and tax timing

Net short- and long-term gains/losses under the applicable rules. Excess net capital losses may offset up to $3,000 of ordinary income annually ($1,500 married filing separately), with unused amounts generally carried forward. The scene illustrates repeated sales of eligible loss-bearing lots and reinvestment in different, reviewed replacement securities. Harvested losses are not investment profit; any tax benefit depends on usable losses, tax rates and individual circumstances. The $1B illustration below states separate aggregate assumptions, rather than adding up the animated bubbles. Lower replacement basis can mean more future taxable gains; eventual sales, tax rates and the investor’s circumstances determine lifetime value.

IRS Topic 409: capital gain/loss treatment ↗MSCI: direct indexing and ETF tax strategies ↗

Tracking error and factors

Tracking error measures the variability of portfolio returns relative to benchmark returns—not a single price gap. The lower bars show 24 synthetic monthly portfolio-minus-index returns calculated from the same two paths. The magnifier shows the latest window of those paths; its price gap is not tracking error. Tracking error would be the sample standard deviation of the monthly return differences, annualized by √12. A lower tracking error does not guarantee better performance. Factor alignment compares systematic exposures such as market, sector, size, value, momentum and quality; stock-specific risk also matters. The factor diagram is schematic, not calibrated market data.

Fidelity: tracking error versus tracking difference ↗MSCI: factor-based portfolio alignment ↗

How to read this scene

The opening fund wrapper compares two ownership structures; it is not an actual ETF redemption. The 48 illustrative bubbles stand for individual holdings: color denotes sector and bubble area approximates weight. After the initial ownership transfer, the cloud remains irregular and active: 11 distinct original holdings are sold and replaced by different securities of the same illustrative sector. No original security is repurchased. Replacement eligibility is assumed after review; matching sector alone does not establish it.

The first 20% of the timeline introduces ownership; the remaining 80% compresses ongoing review, harvesting, reinvestment and weight adjustment. This does not prescribe a trading frequency. Animated positions are schematic, weights are normalized, and movements are not actual orders or a live solver trace. The three factor tracks show schematic alignment toward index targets. They illustrate an optimization objective, not a live PRISM solve or a measured risk result. A real optimizer balances tax value, tracking risk and trading costs, subject to tax-lot eligibility, the client’s restrictions, cash and gain budgets, concentration, liquidity and trade constraints. The return paths are synthetic before taxes and fees; their past never changes when a loss is harvested. The lens and return-difference bars use the same observations as the main chart.

Production inputs and limits

A production workflow also needs verified lots and holding periods, gain budgets, cash flows, client exclusions, concentration and sector limits, liquidity, minimum trades, spreads, market impact, management fees, benchmark changes, corporate actions and household coordination. Hard rules restrict feasible trades; softer preferences enter the objective. Specific lot selection and discrete decisions may require mixed-integer or staged optimization around a continuous risk model. Direct indexing introduces costs and operational complexity; it is not inherently superior to a low-cost fund, and harvesting opportunities can diminish as embedded gains accumulate.

Schwab: implementation, fees and suitability ↗Schwab: benefits and limitations ↗

$1B scale illustration: $1B × assumed 1% eligible harvested losses = $10M. At an assumed 20% applicable tax rate, and with sufficient matching taxable gains in the same taxpayer accounts, $10M × 20% = $2M potential current tax reduction. This is neither a forecast nor a claimed client result. Losses cannot be transferred between unrelated clients. Replacement eligibility and wash-sale screening are assumed; fees, future realization taxes and basis effects may reduce or defer the economic benefit. Harvesting does not erase the underlying investment loss. IRS capital gains and losses.
04 / THE COMPUTATIONAL EVIDENCE

Performance you can
interrogate.

MODEL DIMENSIONS
Tax-lot / TLHSECONDS · LOWER IS BETTER
Recorded-time animation. Not a live solve.
REPORTED MATCHED RESULT
26.67×

vs tuned Gurobi · common portfolio benchmark.

PRISM GPU0.314 s
QUALITY GATESReported pass

S1 · Supplied internal summary. Missing raw timings stay missing. Guarded references never become an inferred speedup.

THE INVESTOR STORY

Choose your route.

Use ← → to navigate, F for fullscreen, M to pause motion, T for theme, N for notes. All controls run locally.
S1 / COMPLETE SUMMARY

Explore the portfolio benchmarks.

PRISM GPU versus tuned Gurobi · quality-capped common portfolio benchmark. Click a cell to inspect it.

Eight portfolio workflows at three sizes
Earlier single-run study, not the current confirmation. Eight displayed portfolio workflows at three sizes. This inspector compares PRISM GPU with tuned Gurobi under the stated quality caps.
EVIDENCE & PROVENANCE

Trace the claim.

Reported measurements, finite models and future strategy are different evidence classes.

No universal solver superiority. No systematic positive after-tax uplift established. No regulatory certification implied by a verification receipt. No live engine, broker connection or client-data upload is part of this briefing.
Benchmark detail and full data

Current charts: eight workflows, three sizes, three repetitions each for PRISM GPU and Gurobi. Earlier seven-solver results remain in the evidence JSON.

S1 · Matched portfolio workloads. One full-workflow observation per executed row. Archive and manifest hashes verified; runs not independently repeated here.

Supporting studies, architecture and full qualifications
Earlier single-run open-reference timing table
100K DIMENSIONS · MATCHED FULL-WORKFLOW COMPARISONSPEEDUP = REFERENCE TIME ÷ GPU TIME
All eight qualified workflows, full-workflow times for four backends, both 100K speedups and scale progression
Time per workload · lower is betterSpeedup versus the named comparison solver
PRESENTER NOTES

The thesis